Beauty in Machine Learning: Purpose and Enlightenment

Abstract

Drawing on empirical research and theories of beauty from psychology, and on philosophical investigations into the role of beauty in scientific research, I argue that aesthetic considerations are valuable to machine research in two modes. First, as a statistical and computational criterion for selecting theories (i.e., models), which leads to quantifiable benefits and appears to be largely automatable. These notions of beauty are largely the same as ``objective'' notions of beauty that are closely tied to human sensory stimulation. Second, somewhat less tangibly and less automatable, as desiderata for our models, methods, and scientific literature. These are manifestations of more complicated notions of beauty that are modulated by experience accumulated over time.

Tasks

Reproductions